On 3/7/11 5:51 PM, Sturla Molden wrote: > Den 07.03.2011 18:28, skrev Christopher Barker: >> 1, 2, 3, 4 >> 5, 6 >> 7, 8, 9, 10, 11, 12 >> 13, 14, 15 >> ... >>
> A ragged array, as implemented in C++, Java or C# is just an array of > arrays (or 'a pointer to an array of pointers'). Sure, but as a rule I don't find direct translation of C++ or Java code to Pyton the best approach ;-) > Basically, that is an > ndarray of ndarrays (or a list of ndarrays, whatever you prefer). > > >>> ra = np.zeros(4, dtype=np.ndarray) > >>> ra[0] = np.array([1,2,3,4]) > >>> ra[1] = np.array([5,6]) > >>> ra[2] = np.array([7,8,9,10,11,12]) > >>> ra[3] = np.array([13,14,15]) > >>> ra > array([[1 2 3 4], [5 6], [ 7 8 9 10 11 12], [13 14 15]], dtype=object) > >>> ra[1][1] > 6 > >>> ra[2][:] > array([ 7, 8, 9, 10, 11, 12]) yup -- or I could use a list to store the rows, which would add the ability to append rows. > Slicing in two dimensions does not work as some might expect: > > >>> ra[:2][:2] > array([[1 2 3 4], [5 6]], dtype=object) yup -- might want to overload indexing to do something smarter about that, though in myuse-case, slicing "vertically" isn't really useful anyway -- the nth element in one row doesn't neccessarily have anything to do with the nth element in another row. However, asside from the slicing syntax issue, what I lose with the approach is the ability to get reasonable performance on operations on the entire array: ra *= 3.3 I"d like that to be numpy-efficient. What I need to grapple with is: 1) Is there a point to trying to build a general purpose ragged array? Or should I jsut build something that satisfies my use-case at hand? 2) What's the balance I need between performance and flexibility? putting the rows in a list give a lot more flexibility, putting it all in one 1-d numpy array could give better performance. NOTE: this looks like it could use a "growable" numpy array, much like one I've written before -- maybe it's time to revive that project and use it here, fixing some performance issues while I'm at it. Thanks for all your ideas, -Chris -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/OR&R (206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception chris.bar...@noaa.gov _______________________________________________ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
